Data center AI chip startup Untether plans to support customer engagements, accelerate second-gen product development.
Canadian data center AI chip startup Untether has completed a $125-million round of funding led by Tracker Capital Management, and co-led by Intel Capital with participation from Canada Pension Plan Investment Board and Radical Ventures. As part of the funding round, Tracker Capital senior advisor Shaygan Kheradpir will join Untether’s board of directors.
“[The funding] is going to support our customer engagements that are ongoing — we’ve had very positive response from the marketplace,” Arun Iyengar, Untether CEO told EE Times. “It’s to make sure that we have the right level of software and customer experience support for the particular neural networks that our customers want to run.”
Iyengar said that Untether’s customer engagements so far have largely been in the enterprise data center segment. Specifically, Untether has engagements in computer vision AI acceleration, including autonomous vehicles, robotics, factory automation and surveillance. Untether engagements in banking and finance span risk analysis, valuation and portfolio balancing, and the company said it is also working with server vendors to integrate its hardware into server products.
Untether’s inference-only chip is a non-Von Neumann at-memory compute design capable of 502 TOPS supporting int8 and int16 arithmetic (read more about Untether’s architecture here). The chips are offered on a four-chip TsunAImi PCIe card delivering 1 PetaOPS for data center inference. Run in a slower mode (720 MHz vs 960 MHz for peak performance), the card enters a sweet spot for power consumption, at which point it can achieve 8 TOPS/W.
Today’s round of funding will also be used to accelerate development of the company’s second-generation chips.
“[Our second-generation product] is going to have more capability than just inference, but it’s still going to be targeting inference,” Iyengar said, noting that inference is still the biggest part of the enterprise AI market.
Untether’s first product is built on 16 nm technology, which the company previously said was down to the ability to achieve a competitive advantage while keeping cost down. Iyengar noted that the company would look at more advanced process nodes for its second-generation chip, but that due to the current shortage of fab capacity, it might choose to stick with a process node below cutting-edge.
“We don’t want to go into the node that everybody else is going to and then find ourselves at the back of the line,” he said.
The company has increased its headcount from 60 to 70 in the last few weeks and Iyengar expects this to grow to 150 over the next 6-9 months. Founded in Canada in 2018, Untether has operations in Toronto and Waterloo and has raised $152 million in funding to date.
This article was originally published on EE Times.
Sally Ward-Foxton covers AI technology and related issues for EETimes.com and all aspects of the European industry for EE Times Europe magazine. Sally has spent more than 15 years writing about the electronics industry from London, UK. She has written for Electronic Design, ECN, Electronic Specifier: Design, Components in Electronics, and many more. She holds a Masters’ degree in Electrical and Electronic Engineering from the University of Cambridge.